Clearly Local ABBYY Cloud OCR SDK

Important

As of 2024-03-15, this repo is now archived, for reference only. For Clearly Local users, please see:

• New demo site at https://cl-tools.deno.dev/pages/easylt-ocr/samples/ocr-sample-editor-poc.html (requires Clearly Local Toolkit account)
• New repo at https://github.com/clearlylocal/easylt-ocr (requires Clearly Local GitHub access)

Usage

OcrSdk class

import { OcrSdk } from 'path/to/OcrSdk.ts'

// credentials obtained from https://cloud.ocrsdk.com/Account/Register
const ocrSdk = new OcrSdk(
	applicationId: Deno.env.get('ABBYY_APPLICATION_ID')!, // e.g. 7ea53f47-8bbc-477b-b17c-989a3184c363
	password: Deno.env.get('ABBYY_PASSWORD')!, // e.g. n6WL0rCFlhU9bDXDri6AQEZV
	serviceUrl: Deno.env.get('ABBYY_SERVICE_URL')!, // e.g. https://cloud-eu.ocrsdk.com/
)

const { txt } = await ocrSdk.ocr(await Deno.readFile('input.jpg'), {
	languages: ['English'],
	exportFormats: ['txt'],
})

await Deno.writeFile('output.txt', new Uint8Array(await txt.arrayBuffer()))

CLI

To run the CLI, note that the relevant ABBYY_APPLICATION_ID, ABBYY_PASSWORD, and ABBYY_SERVICE_URL must be available as environment variables.

# view help
deno task cli --help
# `convert` command, specifying output formats (default "txt")
deno task cli convert path/to/image.jpg -o txt -o xml
# `html`/`json` commands
deno task cli html path/to/image.jpg
deno task cli json path/to/image.jpg
# specify languages (default "English")
deno task cli json path/to/image.jpg -l ChinesePRC -l English

Files

• src/
  • core/
    • OcrSdk.ts: The OcrSdk class, with various methods for interacting with the ABBYY Cloud OCR API. Loosely based on ABBYY's sample JS code, but with the following changes:
      • Rewritten in modern JS/TS with ES6 classes etc.
      • Promise-based API instead of callbacks
      • Calls the v2 (JSON) Cloud OCR API, not the v1 (XML) API
      • Methods now deal with raw binary data rather than file paths (file reading/writing is left up to calling code)
      • Exposes a single ocr method to OCR an image and return the output file binary in the requested format
      • Zero external dependencies
    • imageMap.ts: The imageMap function, for converting XML output to an image map that can be rendered in HTML etc.
    • prettifyXml.ts: The prettifyXml function, for pretty-printing XML output while preserving significant whitespace
    • types.ts: TypeScript types and lists of values for interacting with OcrSdk
  • cli/
    • functions/
      • convertImage.ts: Use OcrSdk's ocr method to get text and XML files of the OCRed content
      • jsonImageMap.ts: Get simplified JSON content from XML using imageMap.ts
      • htmlImageMap.ts: Get XML content, then convert it to a searchable HTML image map based on the text data and coordinates of each line
    • main.ts: CLI app to run the various functions
• samples/
  • ocr-sample.jpg: An example input image file
  • ocr-sample-result.txt: Text file result of running convertImage.ts on ocr-sample.jpg
  • ocr-sample-result.xml: XML result of running convertImage.ts on ocr-sample.jpg
  • ocr-sample-image-map.html: HTML image map generated by running htmlImageMap.ts on ocr-sample.jpg
  • ocr-sample-image-map.json: JSON file generated by running jsonImageMap.ts on ocr-sample.jpg